Re: [Linuxptp-users] PHYs supporting HW timestamping

[Arthur D. <art...@cy...>]

I quick update. I got the DP83640 up and running as PTP-slave with my Cyclone V SoC.
Kernel 4.1.22 did not work for me, I had to change to 4.18 and do the following change in stmmac_main.c in function stmmac_ioctl:
--
case SIOCSHWTSTAMP:
      ret = phy_mii_ioctl(dev->phydev, rq, cmd);
      //ret = stmmac_hwtstamp_ioctl(dev, rq); 
     break;
--

I used the ptp-mmi-tool to enable a 1PPS signal out of the DP83640: 
In ptp-mmi-tool.c I replaced TWOTO32 with TWOTO16 :
//#define TWOTO32   4294967296
#define TWOTO16   65536
to get correct time values when enabling 1PPS.

Now I have a master using MAC-timestamp and a slave using PHY-timestamping.
However, the 1PPS accuracy/jitter is the same (180ns measured over 15min) as when using MAC timestamp for both master and slave.

BR
AD


Sent: Thursday, January 24, 2019 at 4:52 PM
From: "Richard Cochran" <ric...@gm...>
To: "Arthur Dent" <art...@cy...>
Cc: "Wolfgang Hennig" <wh...@xi...>, lin...@li...
Subject: Re: [Linuxptp-users] PHYs supporting HW timestamping
On Thu, Jan 24, 2019 at 03:22:58PM +0100, Arthur Dent wrote:
> I addition to /dev/ptp0 I also have/dev/ptp1 (for the MAC). Can this confuse the system? I read somewhere that the MACs PHC must be disabled...? 

Yes, that is right. Disable the MAC's PHC driver or rip out its code.

Thanks,
Richard